When analyzing the waveform which is outputted from an AD converter or a DA converter, it is not possible to carry out an exact analysis if the noises included in the waveforms are removed. As sources generating noises, there are noises which are mixed into from the power supply or ground, or noises which are possessed by an LSI tester or a measuring apparatus itself.
In order to remove noises, there is a method of performing a filtering processing with using a low pass filter for analog signals while using a filter for digital signals. In these methods, however, it is required to appropriately select a low pass filter or a filter in accordance with an AD converter or a DA converter. In addition, when outputs of plural AD converters or plural DA converters are switched to be outputted to a signal line, the low pass filters or the filters also have to be provided in plural, thereby resulting in complicated circuits.
As another method for removing noises, there is a method of taking a mean of the sampled data. This method makes it possible to cancel random noises which are included in the sampled data and to obtain waveform data having high precision by taking a mean of the sampled data. Further, it is possible to eliminate or minimize the noise reduction circuit such as the above-described low pass filter or filter for reducing noises, and thereby it is advantageously possible to simplify the circuit or equipment that is required for performing an analysis, and further to enhance the reliability of the entire analysis equipment as well as to reduce the cost thereof.
This conventional noise reduction method using taking means will be described with reference to the drawings.
In the conventional method of removing noises shown in FIG. 13, the same waveform data are outputted plural times and then the sampling is performed.
For example, as shown in FIG. 13(a), digital data for four cycles are inputted to the sampling circuit, and sampling is carried out once for 1/n cycle as shown in FIG. 13(b). Then, the sampled data are added together for the same phase sampling points, and the results are stored in a memory. That is, the addition result of data a1, a1′, a1″, a1′″ is stored in a memory as a first phase sampled data, the addition result of data a2, a2′, a2″, a2′″ is stored as a second phase sampled data, and the addition result of data an, an′, an″, an′″ is stored as an n-th phase sampled data. Then, after the sampling of digital data for four cycles are completed, taking a mean processing is carried out by software for respective sampled data corresponding to the respective input data. Thereby, a digital data that is obtained by taking a mean for four cycles is obtained.
As another method of performing noise removal employing a taking a mean processing, there is a method in which the same waveform data are inputted plural times, output data which are obtained by performing sampling at the same timings in the respective waveform data are added together to be successively stored in a memory, and after the sampling is completed, an addition mean is operated by an operator to be outputted. Thus, a part of the addition and outputting processing among the taking a mean processing is carried out while performing sampling, thereby reducing the calculation time (for example, Patent Reference 1).    Patent document 1: Japanese Published Patent Application No. Hei. 1-156682 (Page 8, FIG. 3)